Technology
The Problem
The computer industry is moving from spending money on processors to spending money on communication… the main barrier to scalability. MPT solves the communication problem for both users of computer clusters and manufacturers of multi-core processors.
Why Others Have Failed
The communication problem has long been recognized, and was in fact first accurately described in Amdahl’s Law. The reason that others have consistently failed to bring a true general purpose solution to the industry is as follows:
- They have focused primarily on point-to-point communication improvements.
- The “software stack,” illustrated below, prevents holistic solutions that allow a cluster of servers to behave as a unified problem-solving tool.
- “One size fits all” communication libraries cannot make efficient use of a particular configuration.
Why MPT is Different
MPT has engaged in eight years of intensive and innovative research and development to produce a landmark new communications technology that is both patented and patent pending. MPT’s technical solution has been tested and vetted in four separate DARPA contracts, and is currently the focus of licensing discussions with major OEMs in the parallel processing industry.
Gene Amdahl was quick to recognize the value of MPT’s breakthrough, declaring it to be, “The ultimate solution in parallel processing,” joining the company’s board of advisors, and becoming a shareholder in the company.
The following key points differentiate MPT’s approach from others in the industry:
- We focus on the entire communication task at the cluster level, and optimize it mathematically (22 patents granted or pending)
- Our rigorous algorithms allow a system to manage itself and adapt to changing requirements, at any scale.
- MPT libraries self-train on a given hardware configuration, so there is no tradeoff between optimality and generality.
